harness-everything 0.32.0 → 0.33.0

package/.claude-plugin/marketplace.json

@@ -10,7 +10,7 @@
       "name": "harness-everything",
       "source": "./",
       "description": "Harness OS: an orchestrated agent operating system — tier routing, circuit breaker guardrails, verification gates, WAL session memory, and self-evolving workspace rules.",
-      "version": "0.32.0",
+      "version": "0.33.0",
       "author": {
         "name": "Miya Daniel"
       },

package/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "harness-everything",
   "description": "Harness OS: an orchestrated agent operating system — tier routing, circuit breaker guardrails, verification gates, WAL session memory, and self-evolving workspace rules.",
-  "version": "0.32.0",
+  "version": "0.33.0",
   "author": {
     "name": "Miya Daniel"
   },
@@ -23,6 +23,7 @@
   ],
   "skills": [
     "./multi-agent-workspace",
+    "./code-review",
     "./environment-detection",
     "./eval-harness",
     "./fable-discipline",

package/CHANGELOG.md

@@ -2,6 +2,16 @@
 
 All notable changes to this project are documented in this file.
 
+## [0.33.0](https://github.com/dyphn1/Harness-everything/compare/v0.32.0...v0.33.0) (2026-10-17)
+
+### Features
+
+* **code-review:** add evidence-tiered independent review skill ([#329](https://github.com/dyphn1/Harness-everything/issues/329)) ([2acb220](https://github.com/dyphn1/Harness-everything/commit/2acb22098a5fc360306b0ed34fa3ddcf21592fea))
+
+### Bug Fixes
+
+* **code-review:** register the code-review binding so review steps can load it ([acb766a](https://github.com/dyphn1/Harness-everything/commit/acb766aea36361262f3536db36069265aa66d871))
+
 ## [0.32.0](https://github.com/dyphn1/Harness-everything/compare/v0.31.3...v0.32.0) (2026-10-17)
 
 ### Features

package/README.md

@@ -184,7 +184,7 @@ The authoritative current matrix is [docs/platform-capabilities.md](docs/platfor
 | **Claude Code** | Native lifecycle hooks (`PreToolUse`, `PostToolUse`, `SessionStart`, `UserPromptSubmit`, `Stop`) | `.claude/settings.json`, `.claude/skills/`, `.claude/agents/` | Semantic workflow contracts are reminder-observed; Rule-of-3 reflection and explicit permission boundaries may block |
 | **OpenCode** | Native plugin module ([`opencode-plugin/index.mjs`](opencode-plugin/index.mjs), install as `harness-enforcement.js`) | `.opencode/plugins/` | **Partial live-host evidence** — project-scope `.js` loading and edit/verification state on OpenCode 1.18.31 (macOS); current workflow/verification behavior is reminder-oriented, with only the third-failure reflection boundary blocking edits |
 | **Codex — general installer path** | Skills + `AGENTS.md` instructions | `AGENTS.md` + repo-scoped `.agents/skills/` | **Instruction/advisory delivery only**; semantic MUST/SHOULD/MAY still applies, without a hard-enforcement claim |
-| **Codex / local OpenAI plugin** | `.codex-plugin` package with session, prompt, supported-tool, subagent, and stop hooks plus 26 canonical skills | `.agents/plugins/marketplace.json` → `plugins/harness-everything/` | **Mechanism-tested local contract observation plus explicit permission/Rule-of-3 boundaries** for the packaged mappings; live host loading remains unverified |
+| **Codex / local OpenAI plugin** | `.codex-plugin` package with session, prompt, supported-tool, subagent, and stop hooks plus 27 canonical skills | `.agents/plugins/marketplace.json` → `plugins/harness-everything/` | **Mechanism-tested local contract observation plus explicit permission/Rule-of-3 boundaries** for the packaged mappings; live host loading remains unverified |
 | **Public OpenAI Skills-only plugin** | Public Skills-only bundle | Generated submission ZIP from `plugins/harness-everything/skills/` | **Skill/workflow behavior only**; no local `.codex-plugin` lifecycle hooks in the public artifact |
 | **Cursor** | Native Project Rules + project skills | `.cursorrules` + `.cursor/skills/` | Advisory only |
 | **GitHub Copilot agent surfaces** | Agent Skills + repository custom instructions | `.github/copilot-instructions.md` + `.github/skills/` | Agent Skills path is documented; no live Harness session or plugin install is verified |
@@ -229,7 +229,7 @@ This repo uses a flat layout (waza/agentskills.io convention). The table below m
 | `.agents/plugins` | **Distribution** | OpenAI/Codex repository marketplace metadata |
 | `plugins/harness-everything` | **Distribution** | Local OpenAI/Codex plugin package plus canonical skill copies |
 | `submission/openai` | **Distribution / Review** | Public OpenAI Skills-only listing/test inputs |
-| `evals` | **Routing Evals** | 26 trigger/routing eval suites (waza format) |
+| `evals` | **Routing Evals** | 27 trigger/routing eval suites (waza format) |
 | `eval-framework` | **Quality Gates** | Negative-control fixtures for consistency/collision gates (not a skill) |
 | `contract-integrity` | **Quality Gates** | ADR→spec→ticket→test→implementation trace audit (Phase 1; not a directly routed skill) |
 | `telemetry` | **Quality Gates** | Local JSONL operational evidence layer with report/benchmark scripts (not a skill) |
@@ -238,6 +238,7 @@ This repo uses a flat layout (waza/agentskills.io convention). The table below m
 | `docs` | **Documentation** | Philosophy, architecture, routing, reflection, platform capabilities, generated repository contract, audit |
 | `references` | **Documentation** | Shared checklists (security, performance, definition-of-done) |
 | `multi-agent-workspace` | **Skill (Tier 3)** | Scaffold a verified multi-agent workspace and select bounded specialists from an external catalog without vendoring the full roster |
+| `code-review` | **Skill (Tier 2)** | Independent, evidence-tiered PR/diff code review (read-only by default) |
 | `environment-detection` | **Foundation** | Preflight: detect OS, shell, package manager |
 | `eval-harness` | **Skill (Tier 2)** | Evaluate agent outputs against rubrics |
 | `fable-discipline` | **Skill (Tier 3)** | Fable execution guardrails when Fable is selected |

package/ci/mechanism-2az-step-type-registry.test.js

@@ -832,6 +832,92 @@ checkCase('Claude hook copy is byte-identical and a tracked Codex runtime includ
   compat.uninstall({ codexHome });
 });
 
+checkCase('registered code-review optional binding activates review steps from source and packaged roots', () => {
+  const requirements = [
+    { id: 'req-review', stepType: 'review', summary: 'Review the change', acceptance: 'findings recorded with evidence tiers', optionalBindings: ['code-review'] },
+    { id: 'req-verify', stepType: 'verification', summary: 'Verify the change', acceptance: 'objective checks pass' },
+  ];
+  for (const leg of [
+    { root: ROOT, sessionId: 'code-review-source-root', expected: path.join(ROOT, 'code-review', 'SKILL.md') },
+    { root: path.join(tempRoot, 'code-review-packaged-root'), sessionId: 'code-review-packaged-root', expected: path.join(tempRoot, 'code-review-packaged-root', 'skills', 'code-review', 'SKILL.md') },
+  ]) {
+    if (leg.root !== ROOT) {
+      fs.mkdirSync(leg.root, { recursive: true });
+      for (const directory of ['hooks', 'harness-everything', 'scripts', 'fable-mode']) {
+        fs.cpSync(path.join(ROOT, directory), path.join(leg.root, directory), { recursive: true });
+      }
+      fs.cpSync(path.join(packageRoot, 'skills'), path.join(leg.root, 'skills'), { recursive: true });
+    }
+    const session = planSession(leg.sessionId, requirements, 'iterative-single', leg.root, FIX_PROMPT, 'author-declared plan with an optional code-review binding', leg.root);
+    assert.strictEqual(session.planned.status, 0, session.planned.stderr);
+    const declared = readJson(session.obligationsFile).requirements[0].optionalBindings.find(item => item.id === 'code-review');
+    assert.strictEqual(declared.availability, 'available', 'the registered optional binding must resolve from the declared runtime root');
+    assert.strictEqual(fs.realpathSync(declared.resolvedPath), fs.realpathSync(leg.expected), 'the optional binding resolves to the canonical code-review skill file');
+    const legController = path.join(leg.root, 'hooks/scripts/workflow-disposition.js');
+    const started = run(legController, ['start', '--session-id', leg.sessionId], leg.root);
+    assert.strictEqual(started.status, 0, started.stderr);
+    const load = run(legController, [
+      'binding', '--session-id', leg.sessionId, '--step-id', 'req-review', '--binding-id', 'code-review',
+      '--disposition', 'loaded', '--evidence', 'Read code-review/SKILL.md and applied its evidence-tier rubric.',
+    ], leg.root);
+    assert.strictEqual(load.status, 0, load.stderr);
+    const loaded = readJson(session.obligationsFile).requirements[0].optionalBindings.find(item => item.id === 'code-review');
+    assert.strictEqual(loaded.status, 'loaded');
+    assert.strictEqual(fs.realpathSync(loaded.resolvedPath), fs.realpathSync(leg.expected), 'the loaded binding keeps its canonical resolved path');
+    const passed = run(legController, [
+      'step', '--session-id', leg.sessionId, '--step-id', 'req-review', '--disposition', 'pass',
+      '--evidence', 'Review findings filed against the diff with evidence tiers.',
+    ], leg.root);
+    assert.strictEqual(passed.status, 0, passed.stderr);
+  }
+
+  const notNeeded = planSession('code-review-not-needed', requirements, 'iterative-single', ROOT, FIX_PROMPT, 'author-declared plan marks code-review not needed');
+  assert.strictEqual(notNeeded.planned.status, 0, notNeeded.planned.stderr);
+  assert.strictEqual(run(controller, ['start', '--session-id', 'code-review-not-needed'], ROOT).status, 0);
+  assert.strictEqual(run(controller, [
+    'binding', '--session-id', 'code-review-not-needed', '--step-id', 'req-review', '--binding-id', 'code-review',
+    '--disposition', 'not-needed', '--evidence', 'Reviewed inline against the agentic-commons review rubric.',
+  ], ROOT).status, 0, 'an optional binding can be marked not-needed');
+  assert.strictEqual(run(controller, [
+    'step', '--session-id', 'code-review-not-needed', '--step-id', 'req-review', '--disposition', 'pass',
+    '--evidence', 'Review completed without the optional skill.',
+  ], ROOT).status, 0, 'marking code-review not-needed still lets the review step pass');
+
+  const unknown = planSession('code-review-unknown-optional', [
+    { id: 'req-review', stepType: 'review', summary: 'Review the change', acceptance: 'findings recorded with evidence tiers', optionalBindings: ['no-such-skill'] },
+    { id: 'req-verify', stepType: 'verification', summary: 'Verify the change', acceptance: 'objective checks pass' },
+  ]);
+  assert.strictEqual(unknown.planned.status, 0, unknown.planned.stderr);
+  const undeclared = readJson(unknown.obligationsFile).requirements[0].optionalBindings.find(item => item.id === 'no-such-skill');
+  assert.strictEqual(undeclared.availability, 'unknown', 'an unregistered binding id stays unknown');
+  assert.strictEqual(run(controller, ['start', '--session-id', 'code-review-unknown-optional'], ROOT).status, 0);
+  const cannotLoad = run(controller, [
+    'binding', '--session-id', 'code-review-unknown-optional', '--step-id', 'req-review', '--binding-id', 'no-such-skill',
+    '--disposition', 'loaded', '--evidence', 'attempted load without a registered path',
+  ], ROOT);
+  assert.strictEqual(cannotLoad.status, 2, cannotLoad.stderr);
+  assert.match(cannotLoad.stderr, /no registered load path/, cannotLoad.stderr);
+  assert.strictEqual(run(controller, [
+    'step', '--session-id', 'code-review-unknown-optional', '--step-id', 'req-review', '--disposition', 'pass',
+    '--evidence', 'premature pass while the unknown optional binding is pending',
+  ], ROOT).status, 2, 'a pending unknown optional binding still blocks a premature step pass');
+  assert.strictEqual(run(controller, [
+    'binding', '--session-id', 'code-review-unknown-optional', '--step-id', 'req-review', '--binding-id', 'no-such-skill',
+    '--disposition', 'not-needed', '--evidence', 'No Harness skill exists under that id.',
+  ], ROOT).status, 0, 'an unknown optional binding is marked not-needed instead');
+  assert.strictEqual(run(controller, [
+    'step', '--session-id', 'code-review-unknown-optional', '--step-id', 'req-review', '--disposition', 'pass',
+    '--evidence', 'Review completed without any optional skill.',
+  ], ROOT).status, 0, 'marking the unknown optional binding not-needed lets the step pass');
+
+  const forbidden = planSession('code-review-review-still-forbids-tdd', [
+    { id: 'req-review', stepType: 'review', summary: 'Review the change', acceptance: 'the change is reviewed', optionalBindings: ['tdd'] },
+    { id: 'req-verify', stepType: 'verification', summary: 'Verify the change', acceptance: 'objective checks pass' },
+  ]);
+  assert.strictEqual(forbidden.planned.status, 2);
+  assert.match(forbidden.planned.stderr, /review steps cannot bind tdd/);
+});
+
 fs.rmSync(tempRoot, { recursive: true, force: true });
 console.log(`\n${failed === 0 ? 'PASS' : 'FAIL'}: Stage A workflow step-type registry (${failed} failures)`);
 process.exit(failed === 0 ? 0 : 1);

package/ci/openai-plugin-package.test.js

@@ -104,7 +104,7 @@ const packagedSkillNames = fs.readdirSync(path.join(PLUGIN, 'skills'), { withFil
   .filter(entry => entry.isDirectory())
   .map(entry => entry.name)
   .sort();
-assert.strictEqual(skillNames.length, 26, 'canonical suite should contain 26 skills');
+assert.strictEqual(skillNames.length, 27, 'canonical suite should contain 27 skills');
 assert.deepStrictEqual(packagedSkillNames, skillNames, 'OpenAI package must contain exactly the canonical skill set');
 for (const entry of SOURCE_MANIFEST.skills) {
   const source = path.resolve(ROOT, entry);

package/ci/release-consistency-check.js

@@ -43,6 +43,22 @@ const releaseRenames = new Map([
   ['create-agent-launcher', 'multi-agent-workspace']
 ]);
 
+function logicalSkillName(skill) {
+  return releaseRenames.get(skill) || skill;
+}
+
+function compareSkillCatalogs(tagSkills, currentSkills) {
+  const releaseLogical = new Set(tagSkills.map(logicalSkillName));
+  const currentLogical = new Set(currentSkills.map(logicalSkillName));
+  const missing = [...releaseLogical].filter(skill => !currentLogical.has(skill)).sort();
+  const additions = [...new Set(currentSkills.filter(skill => !releaseLogical.has(logicalSkillName(skill))))].sort();
+  return { missing, additions, exitCode: missing.length ? 1 : 0 };
+}
+
+function formatAdditions(additions) {
+  return additions.length ? `Additions since release tag: ${additions.join(', ')}` : '';
+}
+
 function git(args) {
   return execFileSync('git', args, { cwd: root, encoding: 'utf8' }).trim();
 }
@@ -90,15 +106,13 @@ function main() {
     process.exit(2);
   }
   const current = new Set(discoverSkills(root));
-  const currentLogical = new Set([...current].map(skill => releaseRenames.get(skill) || skill));
-  const releaseLogical = new Set(tagRoot.map(skill => releaseRenames.get(skill) || skill));
-  const missing = [...releaseLogical].filter(skill => !currentLogical.has(skill));
-  const extra = [...currentLogical].filter(skill => !releaseLogical.has(skill));
+  const comparison = compareSkillCatalogs(tagRoot, [...current]);
   console.log(`Release ${tag}: ${tagRoot.length} skill(s); current tree: ${current.size} skill(s).`);
-  if (missing.length || extra.length) {
-    if (missing.length) console.error(`Missing from current tree: ${missing.join(', ')}`);
-    if (extra.length) console.error(`Not present in release tag: ${extra.join(', ')}`);
-    process.exit(1);
+  const additionsLine = formatAdditions(comparison.additions);
+  if (additionsLine) console.log(additionsLine);
+  if (comparison.exitCode !== 0) {
+    console.error(`Missing from current tree: ${comparison.missing.join(', ')}`);
+    process.exit(comparison.exitCode);
   }
 
   let references = 0;
@@ -140,7 +154,10 @@ function main() {
     console.error(`CURRENT RELEASE CHECK FAILED: ${currentReferences.failures.length} dangling reference(s).`);
     process.exit(1);
   }
-  console.log(`RELEASE SKILL CATALOG PASSED for ${tag}; current source has the same logical skill catalog (${tagRoot.length} release entries, ${current.size} current directories after merges).`);
+  const additionLabel = comparison.additions.length === 1 ? 'addition' : 'additions';
+  console.log(`RELEASE SKILL BASELINE PASSED for ${tag}: all ${tagRoot.length} release-tag skills remain present; current source has ${current.size} skill(s), including ${comparison.additions.length} ${additionLabel} not yet in the tag.`);
 }
 
 if (require.main === module) main();
+
+module.exports = { compareSkillCatalogs, formatAdditions };

package/ci/release-consistency-check.test.js

@@ -0,0 +1,30 @@
+'use strict';
+
+const assert = require('assert');
+const { compareSkillCatalogs, formatAdditions } = require('./release-consistency-check');
+
+const addedSkill = compareSkillCatalogs(
+  ['existing-skill'],
+  ['code-review', 'existing-skill']
+);
+assert.deepStrictEqual(addedSkill.missing, [], 'a release-tag skill must remain present');
+assert.deepStrictEqual(addedSkill.additions, ['code-review']);
+assert.strictEqual(addedSkill.exitCode, 0, 'an extra current skill is informational');
+assert.strictEqual(formatAdditions(addedSkill.additions), 'Additions since release tag: code-review');
+
+const missingSkill = compareSkillCatalogs(
+  ['existing-skill', 'removed-skill'],
+  ['existing-skill']
+);
+assert.deepStrictEqual(missingSkill.missing, ['removed-skill']);
+assert.strictEqual(missingSkill.exitCode, 1, 'a released skill missing from current source must fail');
+
+const renamedSkill = compareSkillCatalogs(
+  ['build-multi-agent-system'],
+  ['multi-agent-workspace']
+);
+assert.deepStrictEqual(renamedSkill.missing, [], 'the established rename must satisfy the released entry');
+assert.deepStrictEqual(renamedSkill.additions, [], 'a rename must not be reported as a new addition');
+assert.strictEqual(renamedSkill.exitCode, 0);
+
+console.log('Release consistency catalog additions, missing-skill failures, and rename mapping verified.');

package/ci/router-workflow-plan.test.js

@@ -27,6 +27,12 @@ function check(condition, message) {
   }
 }
 
+function reviewBindingIds(run) {
+  const text = `${run.result.stdout || ''}\n${run.result.stderr || ''}`;
+  const line = text.match(/Candidate step binding ids: ([^\r\n]+)/);
+  return line ? line[1].split(',').map(id => id.trim()).filter(Boolean) : [];
+}
+
 function runTier(prompt, options = {}) {
   const contractPath = path.join(os.tmpdir(), `harness-router-test-${process.pid}-${Math.random().toString(16).slice(2)}.json`);
   const context = options.context || null;
@@ -91,6 +97,36 @@ if (validContract(iterative, 'iterative-single')) {
   check(!plan.requiredInvariants.includes('tdd'), 'binding ids remain step-local rather than workflow invariants');
 }
 
+const reviewPositives = [
+  'Review PR #412 before we merge it',
+  'Review pull request #412',
+  'Review this PR before merge',
+  'Review the diff',
+  'Review the implementation',
+  'Re-review the fix at the new HEAD',
+];
+const reviewNegatives = [
+  'review project plan',
+  'review proposal',
+  'review process',
+  'preview pr environment',
+  'Review this proposal',
+  'Review the project plan',
+  'Review my process',
+  'Re-review the problem statement',
+  'Review my implementation plan, ask one question at a time',
+  'Preview the diff rendering',
+  'Preview PR #412 environment',
+];
+for (const prompt of reviewPositives) {
+  const run = runTier(prompt);
+  check(reviewBindingIds(run).includes('code-review'), `review signal: ${JSON.stringify(prompt)} yields the code-review candidate`);
+}
+for (const prompt of reviewNegatives) {
+  const run = runTier(prompt);
+  check(!reviewBindingIds(run).includes('code-review'), `review signal: ${JSON.stringify(prompt)} stays code-review-free`);
+}
+
 const memoryPersist = runTier('Persist this lesson as memory after resolving the checkout regression.');
 if (validContract(memoryPersist, 'self-evolve persistence')) {
   const plan = memoryPersist.contract.workflowPlan;

package/code-review/SKILL.md

@@ -0,0 +1,41 @@
+---
+name: code-review
+description: "Independently review a PR, diff or finished implementation for correctness, contract and negative-path defects, with source-anchored findings labelled executed, source-traced or inconclusive; read-only by default. USE FOR: \"review this PR before merge\", \"re-review the fix at the new HEAD\". DO NOT USE FOR: \"security threat modeling\", \"implementing or fixing the code\"."
+license: Apache-2.0
+metadata:
+  author: Miya Daniel
+  version: 0.33.0
+---
+
+# Code Review
+
+Independently review completed changes for correctness, contract violations, and negative-path defects. Default to read-only.
+
+## 📋 Skill Contract
+
+| Component | Specification |
+| :--- | :--- |
+| **Trigger / Input** | Requested review of a PR, diff, implementation, or revised HEAD. |
+| **Expected Output** | Source-anchored findings and an evidence-qualified summary. |
+| **State Mutations** | None unless the user explicitly authorizes an action. |
+| **Enforcement Gate** | Apply the checklist in `<this-skill-dir>/references/review-playbook.md`. |
+
+## USE FOR:
+- Review a PR / diff / finished implementation before merge
+- Re-review a revised HEAD: prior-finding disposition + new risks
+
+## DO NOT USE FOR:
+- Security-only threat modeling / secrets / OWASP (use `security-review`)
+- Design or plan stress-testing (use `grill-me`); running gates (use `verification-loop`); writing the fix (use `tdd`)
+
+## Workflow
+
+1. Pin repository, contract, base/head, changed scope, and CI state; inspect relevant callers and consumers.
+2. Independently trace behavior and test applicable counterexamples. Separate severity from safety conditions.
+3. Report only supported findings with evidence tier, smallest correction, regression, and positive control. Use `<this-skill-dir>/templates/finding.template.md`.
+4. On re-review, disposition every prior finding at the new HEAD, then inspect for regressions.
+5. Keep review read-only unless explicitly authorized. Do not claim “bug-free.”
+
+In the interim, bind explicitly on an existing review step: `{"stepType":"review","optionalBindings":["code-review"]}`.
+
+Deep dive: <this-skill-dir>/references/review-playbook.md

package/code-review/references/review-playbook.md

@@ -0,0 +1,40 @@
+# Review Playbook
+
+Use this checklist for the applicable risks; do not force irrelevant categories.
+
+## Establish scope and evidence
+
+- [ ] Pin repository, PR, base/head SHA, user requirement or PRD/ADR, changed files, relevant callers/callees, and current CI state. Avoid unrelated checkout areas.
+- [ ] Make an independent pass. Treat author explanations and tests as leads, not proof; derive expected behavior from contracts and source.
+- [ ] Label evidence honestly: **executed/reproduced** (command, result, environment, SHA); **source-traced** (concrete path and preconditions); **inconclusive** (missing facts or ambiguous semantics; do not present as a proven blocker).
+- [ ] Inventory user-facing entry points and configuration handoffs (CLI flags, cwd, path spellings, config loading). Exercise them end-to-end, not only the recently edited internals.
+- [ ] Reproduce a candidate finding against the PR base before classifying it as new. Pre-existing defects are follow-ups, not new blockers.
+
+## Try applicable counterexamples
+
+- [ ] Missing, null, malformed input; old persisted state; base-to-candidate transitions; tenant/scope boundaries.
+- [ ] Missing or broken dependencies: enumerate every entry point, including help, recovery, and stop paths; inspect fallback behavior.
+- [ ] OS/shell differences; concurrency, crash, retry; user paths and symlinks; API/migration compatibility; downstream or generated views.
+- [ ] Follow each important diagnostic/fact through every consumer and ancestor. Check whether required facts are computed before derived authority, indexes, or output.
+- [ ] Compare new rules with previously accepted rules and controls. Separate display severity from safety prerequisites.
+
+## Check tests and output
+
+- [ ] Ask whether each test still passes if the behavior is disabled, bypassed, or broken (mutation thought experiment). Confirm the test is included in CI.
+- [ ] When output text changes, compare the intended changed lines with literal templates; do not exclude changed lines from assertions.
+- [ ] Dedupe findings. Use P1 for severe merge blockers, P2 for meaningful correctness/contract defects, P3 for low-risk issues; avoid speculation.
+- [ ] For each finding, include exact path:line or diff anchor at HEAD SHA, trigger, observed behavior, violated contract and impact, evidence tier, smallest fix, negative regression, positive control, and disposition.
+
+## Re-review and conclude
+
+- [ ] Pin the new HEAD. Disposition every prior finding separately: resolved, still open, superseded, or inconclusive. Then inspect fix-induced side effects and new changes.
+- [ ] Stay read-only unless the user/workflow explicitly authorizes action. Do not edit, comment, label, or merge on review authority alone.
+- [ ] Summarize scope/SHA, CI, covered and unexamined risks, fixed/open/new counts, and evidence-grounded merge recommendation. “No actionable findings in examined scope” does not mean “bug-free.” Distinguish repository tests from live-host evidence.
+
+## Per-finding format
+
+`[P1/P2/P3] Title` — `path:line @ <HEAD SHA>`
+
+**Trigger** → **behavior** → **contract violation / impact**. **Evidence:** executed / source-traced / inconclusive. **Minimal correction:** … **Regression:** negative case …; positive control … **Disposition:** blocker / follow-up / owner decision.
+
+Use `<this-skill-dir>/templates/finding.template.md` for the complete findings and summary form.

package/code-review/templates/finding.template.md

@@ -0,0 +1,33 @@
+# Code Review — <PR or change>
+
+## Scope and evidence
+
+- Repository / PR:
+- Base SHA → reviewed HEAD SHA:
+- Requirement / PRD / ADR:
+- Changed files and relevant callers/consumers:
+- CI state and commands observed:
+- Counterexample categories covered:
+- Risks not examined / evidence limits:
+
+## Findings
+
+### [P1/P2/P3] <Short, specific title>
+
+- **Location:** `<path>:<line or diff anchor> @ <HEAD SHA>`
+- **Trigger:** <minimal input, state transition, or interleaving>
+- **Behavior → impact:** <observed or source-traced behavior and violated contract>
+- **Evidence tier:** <executed/reproduced | source-traced | inconclusive; include exact command/result or preconditions>
+- **Minimal correction:** <smallest feasible fix>
+- **Regression:** Negative case: <case that fails before fix>; positive control: <valid behavior that remains valid>
+- **Disposition:** <blocker | follow-up | owner decision; for re-review, resolved | still open | superseded | inconclusive>
+
+<!-- Duplicate this finding section as needed. Remove headings for which there are no findings. -->
+
+## Summary
+
+- **Scope / reviewed SHA:**
+- **CI evidence:**
+- **Covered / unexamined risks:**
+- **Findings:** <fixed / open / new counts on re-review>
+- **Conclusion:** <merge recommendation grounded in evidence; “no actionable findings in examined scope” is not a claim of bug-free code>

package/docs/openai-plugin.md

@@ -74,7 +74,7 @@ The local package path is distinct from the general `--codex` installer. The gen
 The package tests prove the following repository-side contract:
 
 - the portable root manifest and `.codex-plugin` compatibility manifest agree on identity, version, interface, and hook entry point;
-- the root `skills/` tree contains exactly the 26 canonical skills;
+- the root `skills/` tree contains exactly the 27 canonical skills;
 - Windows and POSIX hook commands resolve through `PLUGIN_ROOT`;
 - session start emits the compact policy and prompt submission invokes invariant-first routing;
 - supported local `Bash` and `apply_patch` calls receive the packaged guards and state tracking;

package/docs/workflows/code-review.md

@@ -0,0 +1,58 @@
+# Workflow: Code Review
+
+> Independently review a PR, diff, or finished implementation for correctness, contract, and negative-path defects — read-only by default, with evidence-tiered, source-anchored findings.
+
+Source of truth: `code-review/SKILL.md`.
+
+Contract summary from SKILL.md — Trigger: requested review of a PR, diff, implementation, or revised HEAD. Output: source-anchored findings and an evidence-qualified summary. USE FOR: review a PR/diff/finished implementation before merge, re-review a revised HEAD (prior-finding disposition + new risks). DO NOT USE FOR: security-only threat modeling/secrets/OWASP (use `security-review`); design or plan stress-testing (use `grill-me`); running gates (use `verification-loop`); writing the fix (use `tdd`).
+
+## 1. Skill Behavior Workflow
+
+```mermaid
+graph TD
+  A[Pinned Repo Principle BB And CI State] --> B[Read Diff And Relevant Callers Callees]
+  B --> C[Map State Trust Fallback Consumers Tests]
+  C --> D[Construct Counterexamples And Negative Paths]
+  D --> E[Verify Findings Against Source And CI]
+  E --> F{Actionable Supported Finding}
+  F -->|yes| G[P1 P2 P3 Severity Plus Evidence Tier Plus Minimal Fix Plus Regression]
+  F -->|no| H[Discard Qualify Or Mark Inconclusive]
+  G --> I[Report Read Only Only Authorized Actions Post]
+  H --> I
+```
+
+## 2. Triggering and Routing Path
+
+```mermaid
+graph TD
+  ReviewReq[Review This PR Or Diff Request] --> Skill[code-review SKILL]
+  ReReviewReq[Re-Review Fix At New HEAD] --> Skill
+  SecurityReq[Threat Model Secrets OWASP] --> NotCodeReview[security-review SKILL]
+  DesignReq[Stress Test Plan Or Design] --> Grm[grill-me SKILL]
+  GateReq[Run Build Test Lint Gates] --> VL[verification-loop SKILL]
+  FixReq[Implement Or Fix] --> Tdd[tdd SKILL]
+```
+
+## 3. Re-Review Cycle
+
+```mermaid
+graph TD
+  New[Fix Pushed New HEAD] --> Pin[Pin New SHA And Diff]
+  Pin --> Prior[Disposition Every Prior Finding Resolved Open Superseded Inconclusive]
+  Prior --> Regress[Inspect Fix For Induced Side Effects And New Changes]
+  Regress --> Sum[Summary Scope SHA CI Fixed Open New Counts Evidence-Grounded Recommendation]
+  Sum -->|fix needed| Fix[Fix New Defects]
+  Fix --> Pin
+```
+
+Template: `code-review/templates/finding.template.md`. Deep dive: `code-review/references/review-playbook.md`.
+
+## 4. Verification Check
+
+- [ ] Repository, contract, base/head SHA, changed scope, and CI state pinned before reviewing
+- [ ] Independent pass completed; author explanations and tests treated as leads, not proof
+- [ ] Applicable counterexamples constructed (missing/malformed input, persisted state, transitions, boundaries, degraded dependencies, OS/shell, concurrency); irrelevant categories not forced
+- [ ] Every reported finding carries exact path:line/diff anchor at HEAD, trigger, observed behavior, violated contract, impact, evidence tier (executed / source-traced / inconclusive), smallest fix, negative regression, and positive control
+- [ ] Prior findings dispositioned at the new HEAD on re-review, plus fix-induced regressions inspected
+- [ ] Review stayed read-only unless explicitly authorized; no edits, comments, labels, or merges on review authority
+- [ ] Summary distinguished "no actionable findings in examined scope" from "bug-free," with test/CI/live-host evidence boundaries stated

package/evals/code-review/eval.yaml

@@ -0,0 +1,26 @@
+name: code-review-routing-eval
+description: Trigger precision eval for the code-review skill.
+skill: code-review
+version: "1.0"
+config:
+  trials_per_task: 1
+  timeout_seconds: 300
+  parallel: false
+  executor: copilot-sdk
+  model: claude-sonnet-4.6
+metrics:
+  - name: task_completion
+    weight: 0.7
+    threshold: 0.8
+    description: Did code-review trigger exactly when it should, and stay dormant for neighboring skills' requests?
+  - name: efficiency
+    weight: 0.3
+    threshold: 0.7
+    description: Did the skill stay within behavior limits?
+graders:
+  - type: behavior
+    name: token-budget
+    config:
+      max_tokens: 1200
+tasks:
+  - "tasks/*.yaml"

package/evals/code-review/tasks/negative-grill-me.yaml

@@ -0,0 +1,16 @@
+id: negative-code-review-002
+name: Plan stress-testing must route to grill-me, not code-review
+description: Stress-testing a design or plan one question at a time is in code-review's DO NOT USE FOR scope and belongs to grill-me; code-review must stay dormant.
+tags:
+  - negative-trigger
+inputs:
+  prompt: "Re-review my plan, asking one question at a time: stress-test the new billing service plan until we agree on where retries live."
+expected:
+  should_trigger: false
+graders:
+  - type: text
+    name: not-contains-keywords
+    config:
+      contains:
+        - "source-anchored"
+        - "source-traced"

package/evals/code-review/tasks/negative-security-review.yaml

@@ -0,0 +1,16 @@
+id: negative-code-review-001
+name: Security threat modeling must route to security-review, not code-review
+description: A security-only threat modeling request is in code-review's DO NOT USE FOR scope and belongs to security-review; code-review must stay dormant.
+tags:
+  - negative-trigger
+inputs:
+  prompt: "Re-review the STRIDE threat model for security only: do a full threat model of the new payments endpoint against the OWASP Top 10, including secrets exposure."
+expected:
+  should_trigger: false
+graders:
+  - type: text
+    name: not-contains-keywords
+    config:
+      contains:
+        - "source-anchored"
+        - "source-traced"

package/evals/code-review/tasks/negative-tdd.yaml

@@ -0,0 +1,16 @@
+id: negative-code-review-004
+name: Implementing the fix must route to tdd, not code-review
+description: Writing the fix test-first is implementation work that belongs to tdd; code-review is read-only and must stay dormant.
+tags:
+  - negative-trigger
+inputs:
+  prompt: "Fix the off-by-one error in the pagination helper. Write a regression test first, then make it pass."
+expected:
+  should_trigger: false
+graders:
+  - type: text
+    name: not-contains-keywords
+    config:
+      contains:
+        - "source-anchored"
+        - "source-traced"

package/evals/code-review/tasks/negative-verification-loop.yaml

@@ -0,0 +1,16 @@
+id: negative-code-review-003
+name: Running the gates must route to verification-loop, not code-review
+description: Running the build, types, lint and test gates is an objective check that belongs to verification-loop; code-review must stay dormant.
+tags:
+  - negative-trigger
+inputs:
+  prompt: "Run the build, type check, lint and the full test suite on this branch and tell me which gates pass."
+expected:
+  should_trigger: false
+graders:
+  - type: text
+    name: not-contains-keywords
+    config:
+      contains:
+        - "source-anchored"
+        - "source-traced"

package/evals/code-review/tasks/positive-pr-review.yaml

@@ -0,0 +1,16 @@
+id: positive-code-review-001
+name: Pre-merge PR review request triggers code-review
+description: "Independently review a PR, diff or finished implementation for correctness, contract and negative-path defects, with source-anchored findings labelled executed, source-traced or inconclusive; read-only by default. USE FOR: \"review this PR before merge\", \"re-review the fix at the new HEAD\". DO NOT USE FOR: \"security threat modeling\", \"implementing or fixing the code\"."
+tags:
+  - positive-trigger
+inputs:
+  prompt: "Review PR #412 before we merge it. Check the diff for correctness and negative-path defects against the contract in the PRD, and label each finding as executed, source-traced or inconclusive."
+expected:
+  should_trigger: true
+graders:
+  - type: text
+    name: contains-keywords
+    config:
+      contains:
+        - "source-traced"
+        - "inconclusive"

package/evals/code-review/tasks/positive-revised-head-rereview.yaml

@@ -0,0 +1,16 @@
+id: positive-code-review-002
+name: Re-review of a revised HEAD triggers code-review
+description: "Independently review a PR, diff or finished implementation for correctness, contract and negative-path defects, with source-anchored findings labelled executed, source-traced or inconclusive; read-only by default. USE FOR: \"review this PR before merge\", \"re-review the fix at the new HEAD\". DO NOT USE FOR: \"security threat modeling\", \"implementing or fixing the code\"."
+tags:
+  - positive-trigger
+inputs:
+  prompt: "The author pushed a new commit to address last round's findings. Re-review the new HEAD: check each prior finding on its own as resolved, still open or superseded, and look for regressions the fix introduced."
+expected:
+  should_trigger: true
+graders:
+  - type: text
+    name: contains-keywords
+    config:
+      contains:
+        - "prior finding"
+        - "HEAD"

package/harness-everything/SKILL.md

@@ -4,7 +4,7 @@ description: "Route software/project work through Harness: classify tier/topolog
 license: Apache-2.0
 metadata:
   author: Miya Daniel
-  version: 0.32.0
+  version: 0.33.0
 ---
 
 # Harness Everything

package/harness-everything/references/skill-registry.md

@@ -18,6 +18,7 @@ Every skill in this repository is reachable from this router. If a task matches
 | `grill-me` | Tier 2 | User asks to be grilled on a plan (pure Q&A). |
 | `improve-codebase-architecture` | Tier 2 | Deep architectural analysis / refactor planning. |
 | `eval-harness` | Tier 2 | Scoring, stress testing, benchmark comparison. |
+| `code-review` | Tier 2 | Explicit review requests of a PR/diff/finished implementation: correctness, contract, and negative-path findings with evidence tiers, read-only by default — not threat modeling (`security-review`), plan interrogation (`grill-me`), or gate running (`verification-loop`). |
 | `verification-loop` | Tier 2/3 gate | Before declaring done or creating a PR. |
 | `security-review` | Tier 2/3 gate | Changes touching auth, input handling, secrets, network boundaries. |
 | `fable-mode` | Tier 3 | v3 staged planning, named-agent delegation, model selection, failable gates, and cold review. |

package/harness-everything/scripts/routing-keywords.json

@@ -611,6 +611,13 @@
       "bindings": [
         "zoom-out"
       ]
+    },
+    {
+      "id": "review",
+      "bindings": [
+        "code-review"
+      ],
+      "regex": "\\b(?:code[- ]review|pr review)\\b|\\b(?:re-)?review\\s+(?:(?:(?:this|the|my)\\s+)?(?:pr\\b|pull request\\b|diff\\b|(?:finished\\s+)?implementation\\b(?!\\s+plan\\b))|the\\s+(?:fix\\b|new\\s+head\\b))|程式碼審查|代码审查"
     }
   ]
 }

package/hooks/scripts/lib/binding-resolver.js

@@ -40,6 +40,7 @@ const PACKAGED_BINDING_PATHS = {
   'git-commit': 'git-commit/SKILL.md',
   'security-review': 'security-review/SKILL.md',
   'review-guidance': 'harness-everything/references/triage-and-tiers.md',
+  'code-review': 'code-review/SKILL.md',
 };
 const SKILL_ID = /^[a-z0-9][a-z0-9-]{0,63}$/;
 

package/opencode-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "harness-enforcement",
   "description": "Repository inventory for the opencode enforcement plugin and its standalone helper",
-  "version": "0.32.0",
+  "version": "0.33.0",
   "runtime": "index.mjs",
   "hooks": {
     "verify": "hooks/verify.js"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "harness-everything",
-  "version": "0.32.0",
+  "version": "0.33.0",
   "description": "Harness OS: An Orchestrated Agent Operating System for AI-driven Development",
   "main": "harness-everything/scripts/bootstrap.js",
   "bin": {
@@ -20,7 +20,7 @@
     "test:consistency": "node ci/consistency-check.js && node ci/doc-capability-consistency.test.js && node ci/repository-contract.test.js && node ci/mechanism-2ao-plugin-sync.test.js && node ci/prompt-contract-regressions.test.js",
     "test:docs:capabilities": "node ci/doc-capability-consistency.test.js",
     "test:references": "node ci/reference-check.js",
-    "test:release": "node ci/release-consistency-check.js && node ci/mechanism-2x-release-automation.test.js && node ci/mechanism-2x-release-commit-message.test.js && node ci/mechanism-2ap-sbom.test.js",
+    "test:release": "node ci/release-consistency-check.js && node ci/release-consistency-check.test.js && node ci/mechanism-2x-release-automation.test.js && node ci/mechanism-2x-release-commit-message.test.js && node ci/mechanism-2ap-sbom.test.js",
     "test:release:semantics": "node ci/semantic-release-analyzer.test.mjs",
     "test:install": "node scripts/verify-install.js",
     "system-one:install": "node harness-everything/scripts/system-one/install.js",

package/plugins/harness-everything/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "harness-everything",
-  "version": "0.32.0",
+  "version": "0.33.0",
   "description": "Invariant-first engineering harness for routing, verification, and failure recovery while preserving agent autonomy.",
   "author": {
     "name": "Miya Daniel",

package/plugins/harness-everything/hooks/scripts/lib/binding-resolver.js

@@ -40,6 +40,7 @@ const PACKAGED_BINDING_PATHS = {
   'git-commit': 'git-commit/SKILL.md',
   'security-review': 'security-review/SKILL.md',
   'review-guidance': 'harness-everything/references/triage-and-tiers.md',
+  'code-review': 'code-review/SKILL.md',
 };
 const SKILL_ID = /^[a-z0-9][a-z0-9-]{0,63}$/;
 

package/plugins/harness-everything/plugin.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
   "name": "harness-everything",
-  "version": "0.32.0",
+  "version": "0.33.0",
   "description": "Invariant-first engineering harness for routing, verification, and failure recovery while preserving agent autonomy.",
   "author": {
     "name": "Miya Daniel",

package/plugins/harness-everything/skills/code-review/SKILL.md

@@ -0,0 +1,41 @@
+---
+name: code-review
+description: "Independently review a PR, diff or finished implementation for correctness, contract and negative-path defects, with source-anchored findings labelled executed, source-traced or inconclusive; read-only by default. USE FOR: \"review this PR before merge\", \"re-review the fix at the new HEAD\". DO NOT USE FOR: \"security threat modeling\", \"implementing or fixing the code\"."
+license: Apache-2.0
+metadata:
+  author: Miya Daniel
+  version: 0.33.0
+---
+
+# Code Review
+
+Independently review completed changes for correctness, contract violations, and negative-path defects. Default to read-only.
+
+## 📋 Skill Contract
+
+| Component | Specification |
+| :--- | :--- |
+| **Trigger / Input** | Requested review of a PR, diff, implementation, or revised HEAD. |
+| **Expected Output** | Source-anchored findings and an evidence-qualified summary. |
+| **State Mutations** | None unless the user explicitly authorizes an action. |
+| **Enforcement Gate** | Apply the checklist in `<this-skill-dir>/references/review-playbook.md`. |
+
+## USE FOR:
+- Review a PR / diff / finished implementation before merge
+- Re-review a revised HEAD: prior-finding disposition + new risks
+
+## DO NOT USE FOR:
+- Security-only threat modeling / secrets / OWASP (use `security-review`)
+- Design or plan stress-testing (use `grill-me`); running gates (use `verification-loop`); writing the fix (use `tdd`)
+
+## Workflow
+
+1. Pin repository, contract, base/head, changed scope, and CI state; inspect relevant callers and consumers.
+2. Independently trace behavior and test applicable counterexamples. Separate severity from safety conditions.
+3. Report only supported findings with evidence tier, smallest correction, regression, and positive control. Use `<this-skill-dir>/templates/finding.template.md`.
+4. On re-review, disposition every prior finding at the new HEAD, then inspect for regressions.
+5. Keep review read-only unless explicitly authorized. Do not claim “bug-free.”
+
+In the interim, bind explicitly on an existing review step: `{"stepType":"review","optionalBindings":["code-review"]}`.
+
+Deep dive: <this-skill-dir>/references/review-playbook.md

package/plugins/harness-everything/skills/code-review/references/review-playbook.md

@@ -0,0 +1,40 @@
+# Review Playbook
+
+Use this checklist for the applicable risks; do not force irrelevant categories.
+
+## Establish scope and evidence
+
+- [ ] Pin repository, PR, base/head SHA, user requirement or PRD/ADR, changed files, relevant callers/callees, and current CI state. Avoid unrelated checkout areas.
+- [ ] Make an independent pass. Treat author explanations and tests as leads, not proof; derive expected behavior from contracts and source.
+- [ ] Label evidence honestly: **executed/reproduced** (command, result, environment, SHA); **source-traced** (concrete path and preconditions); **inconclusive** (missing facts or ambiguous semantics; do not present as a proven blocker).
+- [ ] Inventory user-facing entry points and configuration handoffs (CLI flags, cwd, path spellings, config loading). Exercise them end-to-end, not only the recently edited internals.
+- [ ] Reproduce a candidate finding against the PR base before classifying it as new. Pre-existing defects are follow-ups, not new blockers.
+
+## Try applicable counterexamples
+
+- [ ] Missing, null, malformed input; old persisted state; base-to-candidate transitions; tenant/scope boundaries.
+- [ ] Missing or broken dependencies: enumerate every entry point, including help, recovery, and stop paths; inspect fallback behavior.
+- [ ] OS/shell differences; concurrency, crash, retry; user paths and symlinks; API/migration compatibility; downstream or generated views.
+- [ ] Follow each important diagnostic/fact through every consumer and ancestor. Check whether required facts are computed before derived authority, indexes, or output.
+- [ ] Compare new rules with previously accepted rules and controls. Separate display severity from safety prerequisites.
+
+## Check tests and output
+
+- [ ] Ask whether each test still passes if the behavior is disabled, bypassed, or broken (mutation thought experiment). Confirm the test is included in CI.
+- [ ] When output text changes, compare the intended changed lines with literal templates; do not exclude changed lines from assertions.
+- [ ] Dedupe findings. Use P1 for severe merge blockers, P2 for meaningful correctness/contract defects, P3 for low-risk issues; avoid speculation.
+- [ ] For each finding, include exact path:line or diff anchor at HEAD SHA, trigger, observed behavior, violated contract and impact, evidence tier, smallest fix, negative regression, positive control, and disposition.
+
+## Re-review and conclude
+
+- [ ] Pin the new HEAD. Disposition every prior finding separately: resolved, still open, superseded, or inconclusive. Then inspect fix-induced side effects and new changes.
+- [ ] Stay read-only unless the user/workflow explicitly authorizes action. Do not edit, comment, label, or merge on review authority alone.
+- [ ] Summarize scope/SHA, CI, covered and unexamined risks, fixed/open/new counts, and evidence-grounded merge recommendation. “No actionable findings in examined scope” does not mean “bug-free.” Distinguish repository tests from live-host evidence.
+
+## Per-finding format
+
+`[P1/P2/P3] Title` — `path:line @ <HEAD SHA>`
+
+**Trigger** → **behavior** → **contract violation / impact**. **Evidence:** executed / source-traced / inconclusive. **Minimal correction:** … **Regression:** negative case …; positive control … **Disposition:** blocker / follow-up / owner decision.
+
+Use `<this-skill-dir>/templates/finding.template.md` for the complete findings and summary form.

package/plugins/harness-everything/skills/code-review/templates/finding.template.md

@@ -0,0 +1,33 @@
+# Code Review — <PR or change>
+
+## Scope and evidence
+
+- Repository / PR:
+- Base SHA → reviewed HEAD SHA:
+- Requirement / PRD / ADR:
+- Changed files and relevant callers/consumers:
+- CI state and commands observed:
+- Counterexample categories covered:
+- Risks not examined / evidence limits:
+
+## Findings
+
+### [P1/P2/P3] <Short, specific title>
+
+- **Location:** `<path>:<line or diff anchor> @ <HEAD SHA>`
+- **Trigger:** <minimal input, state transition, or interleaving>
+- **Behavior → impact:** <observed or source-traced behavior and violated contract>
+- **Evidence tier:** <executed/reproduced | source-traced | inconclusive; include exact command/result or preconditions>
+- **Minimal correction:** <smallest feasible fix>
+- **Regression:** Negative case: <case that fails before fix>; positive control: <valid behavior that remains valid>
+- **Disposition:** <blocker | follow-up | owner decision; for re-review, resolved | still open | superseded | inconclusive>
+
+<!-- Duplicate this finding section as needed. Remove headings for which there are no findings. -->
+
+## Summary
+
+- **Scope / reviewed SHA:**
+- **CI evidence:**
+- **Covered / unexamined risks:**
+- **Findings:** <fixed / open / new counts on re-review>
+- **Conclusion:** <merge recommendation grounded in evidence; “no actionable findings in examined scope” is not a claim of bug-free code>

package/plugins/harness-everything/skills/harness-everything/SKILL.md

@@ -4,7 +4,7 @@ description: "Route software/project work through Harness: classify tier/topolog
 license: Apache-2.0
 metadata:
   author: Miya Daniel
-  version: 0.32.0
+  version: 0.33.0
 ---
 
 # Harness Everything

package/plugins/harness-everything/skills/harness-everything/references/skill-registry.md

@@ -18,6 +18,7 @@ Every skill in this repository is reachable from this router. If a task matches
 | `grill-me` | Tier 2 | User asks to be grilled on a plan (pure Q&A). |
 | `improve-codebase-architecture` | Tier 2 | Deep architectural analysis / refactor planning. |
 | `eval-harness` | Tier 2 | Scoring, stress testing, benchmark comparison. |
+| `code-review` | Tier 2 | Explicit review requests of a PR/diff/finished implementation: correctness, contract, and negative-path findings with evidence tiers, read-only by default — not threat modeling (`security-review`), plan interrogation (`grill-me`), or gate running (`verification-loop`). |
 | `verification-loop` | Tier 2/3 gate | Before declaring done or creating a PR. |
 | `security-review` | Tier 2/3 gate | Changes touching auth, input handling, secrets, network boundaries. |
 | `fable-mode` | Tier 3 | v3 staged planning, named-agent delegation, model selection, failable gates, and cold review. |

package/plugins/harness-everything/skills/harness-everything/scripts/routing-keywords.json

@@ -611,6 +611,13 @@
       "bindings": [
         "zoom-out"
       ]
+    },
+    {
+      "id": "review",
+      "bindings": [
+        "code-review"
+      ],
+      "regex": "\\b(?:code[- ]review|pr review)\\b|\\b(?:re-)?review\\s+(?:(?:(?:this|the|my)\\s+)?(?:pr\\b|pull request\\b|diff\\b|(?:finished\\s+)?implementation\\b(?!\\s+plan\\b))|the\\s+(?:fix\\b|new\\s+head\\b))|程式碼審查|代码审查"
     }
   ]
 }

package/sbom.cdx.json

@@ -2,23 +2,23 @@
   "$schema": "https://cyclonedx.org/schema/bom-1.7.schema.json",
   "bomFormat": "CycloneDX",
   "specVersion": "1.7",
-  "serialNumber": "urn:uuid:10a4f889-cc27-520f-b5d1-1a628d22d914",
+  "serialNumber": "urn:uuid:52f0d0e2-1bad-500d-83b5-0965e29f6e5d",
   "version": 1,
   "metadata": {
-    "timestamp": "2026-10-17T13:06:24.281Z",
+    "timestamp": "2026-10-17T15:47:14.418Z",
     "tools": [
       {
         "vendor": "Harness Everything",
         "name": "generate-sbom.js",
-        "version": "0.32.0"
+        "version": "0.33.0"
       }
     ],
     "component": {
       "type": "application",
-      "bom-ref": "pkg:npm/harness-everything@0.32.0",
+      "bom-ref": "pkg:npm/harness-everything@0.33.0",
       "name": "harness-everything",
-      "version": "0.32.0",
-      "purl": "pkg:npm/harness-everything@0.32.0"
+      "version": "0.33.0",
+      "purl": "pkg:npm/harness-everything@0.33.0"
     }
   },
   "components": [
@@ -1999,7 +1999,7 @@
       ]
     },
     {
-      "ref": "pkg:npm/harness-everything@0.32.0",
+      "ref": "pkg:npm/harness-everything@0.33.0",
       "dependsOn": [
         "pkg:npm/%40commitlint/cli@19.8.1",
         "pkg:npm/%40commitlint/config-conventional@19.8.1",